AI cloaking is the practice of serving one version of a page to an AI crawler and a different one to a human visitor. It's being pitched by a growing number of AI visibility tools as a shortcut to getting cited. It isn't new and it isn't safe. It's the same penalty search engines have enforced for years, wearing new branding.
Key Takeaways |
|---|
The Old Penalty Just Got a New Name
Serving one version of a page to a crawler and a different one to a person is not a new idea. It has a name, cloaking, and it has been against the rules for as long as most of us have worked in search.
What's new is the packaging. A handful of AI visibility vendors are now pitching bot-specific content as a necessary step for getting cited by ChatGPT, Perplexity, or Google's AI answers. The pitch sounds reasonable on the surface: complex sites are hard for bots to parse, so give the bot a simpler version. The mechanics underneath haven't changed. Neither has the risk.
What This Looks Like When the Bot Is an AI Model
The AI version of this pitch usually looks like serving a stripped-down or keyword-dense page to an AI crawler while showing a JavaScript-heavy experience to actual visitors, framed as helping the bot “read” the site.
Most large language models handle standard HTML without much trouble. Maintaining a second version of a page doesn't solve a real technical problem for most sites. It mainly recreates an old risk under new branding, and it doubles the crawl budget a site has to spend for no real gain.
Why It's Riskier Now, Not Less
Google crawls AI Overviews and AI Mode using the same indexing pipeline it uses for core search. A brand caught cloaking doesn't just lose a spot in an AI answer. It risks losing traditional visibility at the same time, in one penalty.
AI crawlers also come back far more often than search crawlers ever did. OpenAI's ChatGPT-User crawler made 3.6x more requests than Googlebot across a sample of 69 sites, per an analysis reported by Search Engine Journal. A mismatch between what a bot sees and what a person sees gets caught faster than it used to.
The temptation is understandable, because the underlying visibility gap is real. Product pages across U.S. retail sites average just 66% on Adobe's AI machine-readability score, meaning roughly a third of that content is effectively invisible to LLMs, according to Adobe. That's a genuine problem. Cloaking is not the fix for it.
What to Do Instead
The fixes that actually hold up are extensions of good technical fundamentals, not workarounds for them:
Server side rendering. Render JavaScript on the server so bots and humans receive the same HTML from the start.
Schema and structured data. Help AI models understand entities and relationships without changing what a person sees.
Clear content structure. Direct headings and answers broken into chunkable sections are inherently easier for AI models to extract.
Consistent facts everywhere. The same product and brand details across every surface an AI system reads.
We put together a fuller technical breakdown of what this looks like for teams building content that's actually meant to be cited: writing content that earns AI citations.
The Real Cost of Getting Caught
Cloaking penalties rarely make headlines. Most brands notice a slow, unexplained decline in visibility long before anyone traces it back to a content mismatch, and by then the damage is already done.
The recovery isn't quick either. Rebuilding trust with a crawler, or with an AI system that has flagged a domain as unreliable, takes months of consistent, verified content, not a quick fix. That's a steep price for a shortcut that was never necessary in the first place.
Where to Start
Pull a sample of your pages and compare what an AI crawler actually fetches against what a human visitor sees. If the two match, you're clear. If they don't, that mismatch is the priority, not a workaround to preserve.
This is exactly the kind of content parity issue that's easy to miss until it costs you visibility. Geckocheck's Brand Intelligence tracks what AI systems actually see on your pages, so you catch the gap before an AI model does.
